Cadence is an AI-first platform that turns your study data into analysis-ready SDTM, ADaM, and TFL deliverables - with a human in the loop at every step.
Built by a team with experience at Gilead Sciences, Genentech, GRAIL, Delfi, and Medpace
Mapping raw clinical data to SDTM, deriving ADaM, and producing TFLs locks study timelines behind a handful of scarce experts. Cadence changes the economics — whether you're a biotech or a CRO.
Reach your first analysis-ready mappings and datasets in days, not weeks.
Your CDISC experts review and approve instead of hand-writing every spec - each one covers more studies.
Traceable artifacts and audit trails are built into the core design.
Five intuitive steps from raw data to validated deliverables — you stay in control.
The TrialPath team will assist in configuring your study context — protocol, CRF metadata, and applied standards — so Cadence is ready to propose mappings and derivations.
Add in your raw datasets as an immutable, checksummed snapshot — the single source every run binds to.
Cadence proposes mapping specs and derivation logic from your protocol, CRF metadata, and applied standards.
Inspect every variable, revise with AI or edit rows directly, and approve. Nothing ships without a human.
Materialize validated SDTM, ADaM, TFL outputs, and other artifacts — each fully traceable back to source for QC and audit.
Cadence is AI-first across the stack — from raw source data to CDISC-aligned SDTM, ADaM, TFLs and other deliverables.
Cadence reads your protocol, CRF, and source datasets to propose domain mappings, source assignments, and the rationale behind each — in dependency order.
SDTM domains — status, class, and next action per domain· 1 / 3
Generate analysis-ready ADaM datasets from approved SDTM, with derivation logic surfaced variable-by-variable so a statistician can verify, not reverse-engineer.
ADaM plan — datasets derived from approved SDTM
Produce TFL outputs wired to their ADaM sources, with full traceability from a number in a table back to the raw column it came from.
TFL plan — outputs wired to their ADaM sources
Every spec, script, and output is AI-drafted and human-approved. Revise in plain English, edit any row directly, and keep a regulatory-grade audit trail throughout.
AI-drafted spec — every variable reviewable before approval· 1 / 2
Ask questions of your study data in natural language — no SQL, no programming. Cadence Explore lets clinical and medical operations teams navigate and interrogate existing datasets directly.
Cadence Explore — natural-language data navigation
Cadence is built for regulated clinical data from day one — your data stays yours, and every action is accountable.
Visit Trust CenterYour study data is never used to train foundation or third-party models. It is used only to produce your deliverables.
Fine-grained roles govern who can view, edit, and approve each study, dataset, and deliverable.
Every spec change, approval, and generation run is logged — who, what, and when — to support inspection readiness.
Deploy in your private VPC to keep clinical data inside your own cloud boundary.
Clinical trials generate the data that brings new medicines to patients — yet turning that data into regulator-ready deliverables still depends on slow, manual, expert-bound work. TrialPath is building the agentic operating system for clinical trial data: AI that does the heavy lifting, with the humans who own the science always in control.
15+ years in clinical trials data across pharma, diagnostics, and medical device. Previously at GRAIL, Gilead Sciences and Medpace.
LinkedIn →
10+ years in clinical software and data engineering. Previously at Delfi Diagnostics, GRAIL, and Genentech.
LinkedIn →The TrialPath team also includes two experienced, founding engineers with experience building software at scale in regulated healthcare environments.
Cadence is built to align with current CDISC standards (SDTM, ADaM, CDASH, Define.xml, Controlled terminologies) and produces deliverables structured to those specifications. CDISC is our primary focus today, but the platform is a general agentic system for clinical trial data and is not limited to CDISC.
Cadence is agentic but not autonomous. AI drafts every spec, script, and output; humans review, revise, and approve, so nothing is finalized without explicit user approval. Every mapping, derivation, and output stays traceable from a final value back to the raw source column it came from, with an audit trail of approvals.
Cadence also includes agentic QC capabilities that help users validate their specs and code - just like a QC programmer would.
No. Your study data is never used to train foundation models, and it is never shared for training purposes. It is used only to produce your deliverables.
Cadence supports deployment in your private VPC so clinical data stays inside your own cloud boundary, with role-based access controls and audit trails for key actions. SOC 2 and HIPAA compliance programs are in progress. We'll walk through your organization's specific security and compliance requirements during a demo.
Cadence ingests common clinical data formats including SAS (.sas7bdat) and CSV, and produces standards-aligned outputs (xpt, csv) ready for downstream QC and submission workflows.
Book a demo and we'll walk you through Cadence on a representative study. We're working with early partners as we finish our first release.
Book a demo. We'll show you SDTM, ADaM, and TFL generation end-to-end — on data that looks like yours.